Is it possible to predict the future of the war in Ukraine? Online forecasting communities think so | Science & Technology

A Russian armed vehicle caught fire in the Ukrainian city of Kharkiv.Serguei BOBOK (AFP)

The day after Russia invaded Ukraine, someone wrote a thank you message on an internet forum. “I just want to say that on February 13 I moved / completely / from Kyiv to Lviv thanks to this prediction thread and the Metaculus estimates. (Still in Lviv but leaving Ukraine later today.) […] Thank you all.”

the message was turned on by a user named “availablegreen”. metaculus, a community dedicated to predicting the future by asking questions like the one that caused this user to leave their town. Will Russia invade Ukraine before 2023? In December, the forecasting community said the probability was 40%. And on February 13, when availablegreen left Kyiv, there was a 60% chance. At that time, the US intelligence services were already saying what would happen. A lot of people didn’t believe it, but availablegreen decided to go ahead anyway.

The changing predictions of whether Russia would invade Ukraine before 2023.

Forecasting platforms are currently experiencing a boom. At metaculus, polymarket, good judgement and insightQuestions are asked about everything. For example about politics: will Emmanuel Macron win the elections in France? Very likely (94%). On the pandemic: will the World Health Organization list a new Covid variant “of concern” in 2022? Likely (74%). Or about catastrophes: what is the danger in London of dying from a nuclear explosion next month? About 24 micro-deaths, 24 options in a million, according to the group of renowned forecasters, which includes Nuño Sempere from Madrid.

Sempere, who writes a newsletter on these topics explains how these platforms work. “Metaculus is a group of people who think these questions are important and that it’s important to have models of the world that can make predictions. Imagine a collaborative community like Wikipedia or Reddit that, instead of writing articles or selecting interesting content, generates investigation and a probability that summarizes it.”

These predictions were popularized by three Pennsylvania teachers, Philip Tetlock, Barbara Mellers and Don Moore, who won a US intelligence-funded competition in 2013. They showed that some people are better at this than others, whom they dubbed “superforecasters,” and that by combining their predictions they could match or surpass the achievements of CIA experts.

But the community that exists now goes even further.

How do you do that?

Forecasters use open sources. They use the information from the Internet, from scientific studies and press releases to public data. as said Sam Freedman: “Anyone on Twitter, if they filter information well, can be better informed about the real-time course of the war than Eisenhower about Korea or LBJ about Vietnam.”

They also know the recipe for better forecasts. First, a certain type of viewpoint is required: quantitative, probabilistic, thrifty, willing to change one’s mind. Second, using aggregation methods is better than a median method (for example, if two people with different information tell you that there is a 50% chance of rain, you should bet that there is a greater than 50% chance of rain). And third, logical but often forgotten: you really have to want to do it right.

How much are they getting right?

Within the community they are not very happy with their progress so far during the war in Ukraine, although I think there is some justification for the fact that they said in January that the invasion was somewhat likely.

I also followed another prognosis of theirs that started out as a failure but soon changed. It was useful for me to organize my journalistic coverage: would Kyiv fall under Russian control before April 1? On the second day of the invasion of Metaculus, they believed it was likely to fall (80%), as did the majority of observers, who expected a rapid Russian advance. But they soon corrected that view.

By day five, the probability of Kyiv coming under Russian control had dropped to 67%, by March 6 it was 37%, and by March 15, two weeks before the deadline, it was down to 10%. Even more interesting, they now believe there is an 80-85% chance that the capital will resist for another two months, until June 1st. Will they be right?

The changing prognosis on whether Russia will control Kyiv by June 1, 2022.

One of Metaculus’ successes has been its predictions during the Covid pandemic, as explained by New York’s Juan Cambeiro. I met him when he was leading the ranking of the best forecasters and now he works for the platform. “Around 2. DecemberMetaculus forecasters successfully predicted that Omicron would quickly overtake Delta.” And also that it has “intrinsically greater transmissibility that would undermine vaccine protection and be even less deadly than Delta.”

Of course, the people of Metaculus are neither oracles nor infallible. They almost never offer absolute predictions. But they’ve been proven right in one key sense: they are very well calibrated. The events they rate as 60% probability occur 60% of the time (more or less), and those with 90% probability occur 90% of the time (more or less).

Why are they doing it?

There are platforms where you can forecast to make money or cryptocurrency, but the main motives seem to be hobby and dedication for now. If the internet has taught us anything, it’s that people can dedicate a lot of time to their interests.

It’s not a game for Sempere. “Sports betting is ridiculous to me and I don’t see the fun,” he says. “They’re structured to create addiction.” He also forecasts in competitive and paid markets, but he finds pros and cons there. “They allow you to invest more effort,” he says. “But working together is more difficult.”

The key is that both Sempere and Cambeiro believe their work is useful and has a lot of potential. We all have to make quick decisions, often under uncertainty. That’s obvious to a mayor or executive, but so is the man who’s closed his bar during the pandemic, or the youngster who’s choosing not to buy a home for fear of a recession. These platforms, Sempere says, can produce probabilities about “how long a quarantine will last or who the next president will be.” They won’t decide for you, but they can shed light and inform your choices. Cambeiro points out that this is already happening. “Many people have made decisions about Covid based on our decisions,” he explains. “Many users and I took precautions before anyone else.”

Could this be the case with availablegreen? I can’t guarantee it’s true, but I’ve spoken to him and it does seem to be the case. He is a young Belarusian living in Kyiv who remembered reading about these platforms in February (“I went to see what the prediction markets said”) when he was already concerned. Trust in her and in what he could see The New York Times and on Twitter he made a decision: to leave the city to live in Lviv, from where he later traveled to Warsaw.


Comments are closed.